- 130 -
CHAPTER 4:
DEVELOPING A MODEL OF MARKET-DRIVING ABILITY IN CORPORATE
ENTREPRENEURSHIP
4.1 INTRODUCTION
“Theory development typically focuses on relationships among theoretical constructs,
placing little emphasis on relationships between construct and measures. In most
cases, constructs are treated as causes of their measures …” (Edwards & Bagozzi,
2000:155).
Considerable attention has been paid over the past 25 years to the scale
development process in order to improve validation of constructs. However, these
procedures are grounded in classical test theory, which assumes that the constructs
cause their measures (reflective perspective) and do not consider that there are
cases in which the indicators cause the latent construct (formative perspective)
(MacKenzie, Podsakoff & Burke Jarvis, 2005:710).
It is observed that most researchers in social sciences assume that indicators of
latent constructs are reflective, despite the fact that formative indicators are
appropriate (Diamantopoulos et al., 2008:1204). Reasons for this situation may be
the convenience factor in analysing models under the reflective view. A number of
programmes are available for the analysis of covariance-based structural equation
modelling (Law & Wong, 1999:156).
However, researchers argue that technical convenience should not guide the
research and the adoption of the type of measurement model (Law & Wong,
1999:159). Furthermore, model misspecification has a serious impact on the
conclusions on and interpretation of models (Diamantopoulos et al., 2008;
MacKenzie et al., 2005; Podsakoff, MacKenzie, Podsakoff & Yeon Lee, 2003).
- 131 -
The purpose of this chapter is threefold. First, it will provide a literature review of
measurement instruments that have been used in past research relating to the
various constructs defined in chapter three.
Second, it will present in a literature review various types of measurement models
and their characteristics. Third, it will develop a model for market-driving ability and
outline propositions.
4.2 LITERATURE REVIEW ON MEASURING INSTRUMENTS
In order to prepare the conceptual model of market-driving ability developed in
chapter three for measurement, an analysis of selected measuring instruments that
have been used in previous research is presented. The following paragraphs outline
the dimensions of each measurement instrument and its reliability and validity.
Reliability is a measure to assess the extent to which measures provide consistent
results (Cooper & Schindler, 2008:292). Reliability can be estimated by internal
consistency, the split-half approach or the test-retest method. Internal consistency is
a predominant measure in research studies. It is measured by Cronbach’s alpha,
which reflects the correlation of the number of items and their average correlation
(Nunnally & Bernstein, 1994:251,254). Cronbach’s alpha can achieve values
between zero and one, whereby the closer the value is to one the more reliable the
scale (Santos, 1999:2). Nunnally (1978 in: Santos, 1999:2) indicates that values at
0.7 represent acceptable reliability.
Reliability is considered as a necessary, but not sufficient, contribution to validity
(Cooper & Schindler, 2008:292). Validity measures comprise measures for external
and internal validity. External validity is concerned with the generalisability of
research results. Internal validity refers to the ability of the research instrument to
measure what it is supposed to measure (Cooper & Schindler, 2008:289). Within
internal validity three types can be distinguished: content validity, construct validity
and criterion-related validity (Cooper & Schindler, 2008:290-292; Nunnally &
Bernstein, 1994:108).
- 132 -
To ensure content validity it is necessary to have a well-formulated plan on how to
measure certain items before the test is applied. The items can also be presented to
a panel of persons to assess the validity of the items (Cooper & Schindler, 2008:290;
Nunnally & Bernstein, 1994:102-103).
Construct validity considers theory and the measurement instrument. Construct
validity considers aspects of convergent and discriminant validity. Convergent validity
can be assessed by correlating the developed scale with other scales that purport to
measure the same construct. Discriminant validity is determined by relating the
developed scale to measures that are supposed to measure different constructs
(Cooper & Schindler, 2008:291-292).
Criterion-related validity considers how well the measures can be used for estimation
or prediction (Cooper & Schindler, 2008:291). To estimate whether a specific item
serves as a valid measure for the scale, the item needs to be correlated with two
groups that are supposed to be different. If no difference between the two groups
regarding the item can be established, concurrent validity is not achieved (Bryman &
Bell, 2007:165).
Most of the presented scales have been developed as a part of a broader study
which often relates the scales to other constructs. However, for the purpose of this
study only the scales that can potentially be used to measure market-driving,
antecedents and consequences of market-driving ability are presented.
4.2.1 Measuring instruments for market driving
In chapter three market driving was outlined as a specific firm behaviour. Certain
abilities such as market sensing, influencing customer preferences and alliance
formation characterise market driving (Barlow Hills & Sarin, 2001, 2003; Ghauri et al.,
2008; Harris & Cai, 2002; Jaworski et al., 2000). However, no measurement
instrument for market driving has been developed so far.
Market sensing has been defined as capturing aspects of forward-looking firm
activities that consider all relevant stakeholders. This aspect is closely related to
- 133 -
influencing customer preferences, which also needs a forward-looking approach, as
well as an information component in order to educate customers about new products
and services (Barlow Hills & Sarin, 2001, 2003; Ghauri et al., 2008; Harris & Cai,
2002; Jaworski et al., 2000).
The following measuring instruments are considered to capture aspects of the
concepts of market sensing and influencing customer preferences.
Narver et al. (2004:336) developed a scale entitled MOPRO (proactive market
orientation) which consists of items that capture customer’s latent needs by
monitoring customer behaviour and exceeding customer expectations. Furthermore,
a proactive market orientation involves leading the customer.
The MOPRO scale consists of 34 items. After exploratory factor analysis was
conducted, 11 items remained with a Cronbach’s alpha of 0.892. In confirmatory
factor analysis it was found that the construct was not unidimensional. Reducing the
items to eight resulted in a unidimensional measure with satisfactory fit indices.
Convergent, divergent and discriminant validity were demonstrated, using the
responsive market orientation scale as a comparison (Narver et al., 2004:339-341).
A similar activity has been described as “scanning intensity” by Barringer and
Bluedorn (1999:423). Environmental scanning refers to activities aimed at learning
about events and trends in the firm’s environment. The information that is obtained
from scanning can be further used to recognise opportunities or reduce uncertainty.
A 12 item scale for scanning intensity was developed, focusing on the efforts taken
towards environmental scanning and the comprehensiveness of the scanning
process (Barringer & Bluedorn, 1999:428). Six items measured the effort towards
scanning and included modified items from Miller and Friesen’s (1982) effort
dedicated towards a scanning scale. The comprehensiveness was also measured by
six items asking for the scanning elements that are used in the firm. A mean score
over the 12 items constituted the scanning intensity (Barringer & Bluedorn,
1999:428).
- 134 -
Reliability was assessed by Cronbach’s alpha, which showed acceptable values
(alpha = 0.83). Discriminant validity was established by exploratory factor analysis.
Two modified items from Miller and Friesen’s (1982) measures did not load high on
any factor. However, for conceptual reasons the two items remained in the scale
(Barringer & Bluedorn, 1999:429-430).
The scales developed by Narver et al. (2004) and Barringer and Bluedorn (1999)
demonstrate high reliability values, with Cronbach’s alpha scores of 0.89 and 0.83
respectively. Hence, items of these scales can be used in further studies. However, it
needs to be considered that the scales have not been assessed on a longitudinal
basis or across countries and industries (Narver et al., 2004:344).
Alliance formation represents the third concept in the market-driving construct.
In the literature strategic alliance is described as a voluntary agreement between
independent firms for exchange, sharing or co-development (Gulati, 1999:397;
Ireland et al., 2002:413; Kale et al., 2000:218; Rothaermel & Deeds, 2006:430).
The use of strategic alliances is most often operationalised as an independent
variable indicating a specific number of alliances in the areas of research and
development, marketing, licensing agreements or cross-licensing. The duration of
strategic alliances is also measured (Deeds & Hill, 1996:48; Dickson & Weaver,
1997:411; Gulati, 1999:405; Hitt, Dacin, Levitas, Arregle & Borza, 2000:457; Kale et
al., 2000:226).
Alliance management capability refers to a firm’s ability to effectively manage
multiple alliances. It is argued that alliance management capability is built through
alliance experience and the alliance type (Rothaermel & Deeds, 2006:430-431).
Alliance type was measured through the number of research and development
alliances the firm had entered. Alliance experience considered the firms alliance
duration. Alliance management capability was measured by the number of alliances
a firm was able to manage productively (Rothaermel & Deeds, 2006:441-442).
- 135 -
Discriminant validity for alliance management capability was established
(Rothaermel & Deeds, 2006:445).
Kale et al. (2000:220) measured the aspects of trust which develops between firms
based on close interaction at a personal level with the construct “relational capital”. A
five item measure was used to capture mutual trust between alliance partners
(Kale et al., 2000:237). Reliability was assessed using Cronbach’s alpha, which was
very satisfactory (alpha = 0.906). Content validity was established by pre-testing the
survey instrument (Kale et al., 2000:226).
Rothaermel and Deeds (2006) and Kale et al. (2000) provide valid aspects for the
measurement of alliance. Rothaermel and Deeds (2006) focus on the management
aspect of alliances, whereas Kale et al. (2000) research the concept of trust that
needs to be present in alliances. Both aspects are considered to be important in the
measurement of alliance formation.
4.2.2 Measuring instruments for entrepreneurship
The following paragraphs present selected studies that developed or replicated
measuring instruments for entrepreneurship and corporate entrepreneurship.
4.2.2.1 Entrepreneurial orientation measuring instruments
The following paragraphs outline several scales for measuring entrepreneurial
constructs.
One of the most widely used measurement instruments in entrepreneurship is the
construct of entrepreneurial orientation, which was originally developed by Miller and
Friesen (1982) and subsequently developed further by Covin and Slevin (1986)
(Kreiser et al., 2002:71).
Entrepreneurial orientation is considered to be a multidimensional construct that
encompasses a firm’s activities regarding innovation, risk-taking and proactiveness
(Miller & Friesen, 1982:7,17-24; Miller, 1983:770-771).
- 136 -
Miller and Friesen (1982:7-10) analysed innovation and risk-taking together with
other strategic and environmental variables. Good construct reliability for innovation
and risk-taking (Cronbach’s alpha = 0.77; 0.91 respectively) was achieved. The
concept of proactiveness was not explicitly included in the study.
Covin and Slevin (1986:629-631) developed a measurement scale of entrepreneurial
behaviour taking into consideration items developed by Miller and Friesen (1982) and
Khandwalla (1977). A total of 10 items was used to measure entrepreneurial
behaviour. Risk-taking consisted of six items that were taken from Khandwalla
(1977). Product innovation consisted of two items adapted from Miller and Friesen
(1982). Proactiveness consisted of two self-constructed items considering Miller and
Friesen’s (1982) description of the concept.
Factorial validity was assessed to determine unidimensionality. The analysis showed
that four items loaded poorly on a single factor. These were items from the risk-taking
dimension. The remaining six items showed acceptable results. Reliability was
assessed using Cronbach’s alpha for the six items, which showed satisfactory results
(alpha = 0.79). Finally an entrepreneurship index was calculated, using the mean
score of the six items (Covin & Slevin, 1986:632-634,638-639).
Considering the fact that four items out of 10 did not represent the entrepreneurship
construct it needs to be questioned whether sufficient validity of the construct has
been established. Furthermore, a validation of the scale compared with other scales
would have been beneficial, to establish more confidence in the entrepreneurship
index.
Since innovation was measured subjectively on the scale by Miller and Friesen
(1982:922), Jennings and Young (1990:54) developed an objective measure of
product innovation and compared their scale with the subjective measures.
The objective measures were developed based on five financial ratios which served
as indicators for product innovation. The subjective measure consisted of three items
developed by Miller and Friesen (1982) to measure innovation activities with regard
to new product development (Jennings & Young, 1990:57).
- 137 -
Construct reliability was demonstrated for the objective and subjective measures
(Cronbach’s alpha = 0.75; 0.92 respectively). Correlation analysis for both measures
was performed and resulted in a significant relationship between the two measures
(Jennings & Young, 1990:58-61). Jennings and Young (1990:62) concluded that
objective and subjective measures of product innovation can be used
interchangeably.
Besides the well-known three concepts of innovation, risk-taking and proactiveness
to measure entrepreneurial orientation, Lumpkin and Dess (1996:140,148)
conceptually defined entrepreneurial orientation consisting of five dimensions: the
three dimensions developed by Miller and Friesen (1982) and also autonomy and
competitive aggressiveness.
In a separate study, Lumpkin and Dess (2001:439,441-443) operationalised
entrepreneurial orientation using four concepts, namely risk-taking, innovation,
proactiveness and competitive aggressiveness. Items for risk-taking, innovation and
proactiveness were taken from scales developed by Khandwalla (1977), Miller (1983)
and Covin and Slevin (1986,1989) and partly adapted. Competitive aggressiveness
was measured by two items. One was taken from Covin and Slevin (1989) and one
was self-constructed. It was found that the four concepts of entrepreneurial
orientation represent distinct factors. Reliability was reported for proactiveness and
competitive aggressiveness, which demonstrated a satisfactory value for
proactiveness (Cronbach’s alpha = 0.79) and a less reliable value for competitive
aggressiveness (Cronbach’s alpha = 0.66).
Although Lumpkin and Dess (2001) demonstrated that entrepreneurial orientation
can be measured by four distinct concepts, the three dimensions developed by Miller
and Friesen (1982) have been most widely used in entrepreneurship and strategic
management research (Kreiser et al., 2002:71).
Knight (1997:215) assessed the scale developed by Khandwalla (1977) and termed it
the ENTRESCALE. Eight items were used in the study. In order to identify whether
entrepreneurial orientation is a unidimensional or multidimensional construct, a factor
analysis was performed, which showed a two-factor structure. All items were
- 138 -
assessed regarding their reliability and an overall Cronbach’s alpha value of 0.834
was established (Knight, 1997:218-219). Knight (1997:219) used structural equation
modelling to assess construct validity. The measurement model fitted the data well,
lending support for construct validity. Construct validity was tested in the form of
discriminant validity, using a correlation analysis correlating the aggregated
ENTRESCALE dimensions with other relevant entrepreneurship measures, which
showed satisfactory results (Knight, 1997:220).
Chadwick et al. (2008:70,76) examined the entrepreneurial orientation scale with
nine items covering innovation and proactiveness. The scale demonstrated good
reliability, with Cronbach’s alpha values at 0.82. The reliability is consistent with
Knight’s (1997) findings. Factor analysis was conducted, which showed a reliable
two-factor structure which was also found in Knight’s (1997) study. Convergent
validity was assessed using correlation analysis between the ENTRESCALE and a
measure of proactiveness developed by Venkatraman (1989; in Chadwick et al.,
2008:75). Support for convergent and nomological validity was found (Chadwick et
al., 2008:75-76).
The studies presented so far consider that entrepreneurial orientation consists of
three dimensions: innovation, risk-taking and proactiveness, which are either present
in an organisation or not. Morris and Sexton (1996), however, developed an
approach which demonstrates that entrepreneurial orientation can be assessed
based on the degree and amount of entrepreneurship that takes place in an
organisation.
Morris and Sexton (1996:5-9) developed the construct of entrepreneurial intensity.
The underlying concepts are innovation, risk-taking and proactiveness. It is argued
that entrepreneurship is a matter of degree and frequency. The degree of
entrepreneurship was measured by items covering the extent of top-management
decision making in an innovative, risk-taking and proactive way. These items were
developed by Miller and Friesen (1983:232) and adapted by other researchers. Next,
the number of new products, services and processes was assessed, by indicating if
they were new to the world, new to the market or modifications or extensions of pre-
existing items.
- 139 -
Satisfactory reliability could be achieved for the degree of entrepreneurship.
Innovativeness demonstrated a Cronbach’s alpha of 0.84, risk-taking was at 0.72 and
proactiveness at 0.67. Regarding frequency of entrepreneurship, absolute numbers
were reported. The entrepreneurial intensity measure was calculated by mean scores
of the degree of entrepreneurship and the average of responses on the frequency
dimension. The measure of entrepreneurial intensity was an equally weighted index
of the combination of degree and frequency (Morris & Sexton, 1996:9-10).
The presented studies on the measurement of entrepreneurial orientation showed
that the construct can be measured reliably by the three concepts of innovation, risk-
taking and proactiveness. Whereas some researchers argue that entrepreneurial
orientation is a unidimensional construct (Covin & Slevin, 1986), others demonstrate
that it is a multidimensional construct consisting of distinct factors (Chadwick et al.,
2008; Knight, 1997; Lumpkin & Dess, 2001; Miller & Friesen, 1982).
George (2011:3) takes the discussion further and argues that entrepreneurial
orientation is not only a multidimensional construct, but that it can also be modelled
as a reflective or formative construct. The differences and implications of reflective
versus formative measurements will be outlined in section 4.3.1.
4.2.2.2 Corporate entrepreneurship measuring instruments
Khandwalla (1977:23,424,637) developed a measurement instrument to assess
corporate design of organisations. Corporate design is influenced by environmental,
strategic, structural and behavioural constructs. Reliability was measured using
Nunnally’s formula for reproducibility. The values are stated in brackets where
applicable.
The environmental construct consisted of variables measuring research and
development activities (n.a.), the rate of innovation (0.76), competitive pressure
(0.56) and the external environment (n.a.) (Khandwalla, 1977:639-642,659).
The strategic construct included variables measuring performance aspirations (0.69),
the organisation’s orientation to diversification (n.a.) and vertical integration (n.a.),
- 140 -
standardisation orientation (0.45), risk-taking (0.53), optimisation of use of resources
(0.80), participation (0.85), flexibility (0.68) and coercion of top management (0.52)
(Khandwalla, 1977:642-650,660-661).
The technological construct consisted of variables capturing capital intensity (n.a.)
and orientation towards mass production (n.a.) (Khandwalla, 1977:650-651,661).
The structural construct included measures of delegation of authority (0.81),
distribution network (n.a.), vertical integration (0.69), divisionalisation (n.a.) and
sophistication of control and information systems (0.80) (Khandwalla, 1977:651-655).
The control of behaviour construct was used to assess management’s activities to
reduce conflict or improve coordination (n.a.) (Khandwalla, 1977:655-656).
Performance was measured with an index of subjective (0.84) as well as an index of
objective performance (n.a.) (Khandwalla, 1977:656-658).
Although values for competitive pressure (0.56), performance aspirations (0.69), risk-
taking (0.53), coercion (0.52) and flexibility (0.68) are below the suggested criterion
of 0.7, they were included in the study since the research was in its early stages and
low values can be accepted in that stage (Khandwalla, 1977:658).
Overall the scale developed by Khandwalla (1977) showed that different aspects of a
corporate management style can be measured reliably.
Zahra (1991:272) developed a measure for corporate entrepreneurship. Zahra states
that his corporate entrepreneurship indicators cover an organisation’s actual
engagement in entrepreneurship, whereas other measures such as the scale
developed by Miller (1983) measure the disposition towards entrepreneurship.
Corporate entrepreneurship was measured by four indicators, a corporate
entrepreneurship index, sales derived from new business lines, sales derived from
new products or brands and an external orientation of corporate entrepreneurship
(Zahra, 1991:271-272).
- 141 -
First, a corporate entrepreneurship index was developed. The index consists of nine
items which showed acceptable reliability (Cronbach’s alpha = 0.86). The items relate
to areas such as supporting and rewarding employees, engagement in innovation,
organisational structure, management support, competitor orientation and
environmental aspects (Zahra, 1991:271,285). The corporate entrepreneurship index
was shown to be valid when correlated with Miller’s (1983) index consisting of the
concepts of innovation, risk-taking and proactiveness. Further, a clear distinction
from Miller’s (1983) index could be shown (Zahra, 1991:271).
The second measure in Zahra’s (1991) model is percentage of sales derived from
new lines of business. The third measure considers percentage of sales derived from
new products or brands. The last measure accounts for external orientation of
corporate entrepreneurship, which was measured by the number of joint ventures the
organisation had participated in the past three years. Reliability for these indicators
was established using objective as well as subjective data (Zahra, 1991:272).
Zahra (1991:278) also demonstrated a positive association of corporate
entrepreneurship with financial performance measures.
Antoncic and Hisrich (2001:495-496) used the ENTRESCALE (Knight, 1997) and the
corporate entrepreneurship scale by Zahra (1991) to further refine the measurement
scales and assess their cross-national validity.
The construct of corporate entrepreneurship consisted of 37 items that measured
four concepts. Reliability in the form of Cronbach’s alpha was assessed for the
Slovenian and the US sample. The values are provided in brackets. The four
concepts consider: new business venturing (0.83/0.51), innovativeness (0.89/0.87),
self-renewal (0.92/0.83) and proactiveness (0.69/0.66). Except for proactiveness, all
concepts showed acceptable levels of reliability. Exploratory factor analysis and
confirmatory factor analysis were conducted for all dimensions and showed
satisfactory results. Convergent and discriminant validity were also established
(Antoncic & Hisrich, 2001:517-518).
Overall, the corporate entrepreneurship construct showed acceptable internal and
external validity regarding the generalisability across the two samples (Antoncic &
- 142 -
Hisrich, 2001:521). These results give confidence for the use of corporate
entrepreneurship items in future international research studies.
Kuratko et al. (1990:54) developed a comprehensive measurement of corporate
entrepreneurship in the “Intrapreneurial Assessment instrument (IAI)”.
The scale was further refined by Hornsby, Montagno and Kuratko (1992 in Hornsby
et al., 1999:12) leading to the “Corporate Entrepreneurship Assessment Instrument
(CEAI)”.
Hornsby et al. (2002:253) applied the Corporate Entrepreneurship Assessment
Instrument (CEAI) to middle managers. The purpose of the study was to assess firm-
internal factors that influence middle management’s participation in corporate
entrepreneurship activities.
The following five concepts were used to measure corporate entrepreneurship:
management support, organisational structure, risk-taking, time availability and
reward and resource availability. The concepts were operationalised with a total of 84
items (Hornsby et al., 2002:263).
Exploratory and confirmatory factor analysis resulted in five factors. Reliability for all
factors was established with Cronbach’s alpha. The values are provided in brackets.
The five factors are: management support (0.92), work discretion (0.86),
rewards/reinforcement (0.75), time availability (0.77) and organisational boundaries
(0.69). Results from a second sample also showed a five-factor model and high
reliability (Hornsby et al., 2002:266-267). Discriminant validity was established for all
five factors (Hornsby et al., 2002:268).
The presented studies by Khandwalla (1977), Zahra (1991) and Hornsby et al. (2002)
demonstrate that corporate entrepreneurship is a diverse construct. Besides the
entrepreneurial orientation constructs of innovation, risk-taking and proactiveness,
other relevant constructs have been considered. All three studies show that
entrepreneurial orientation can reliably be measured by internal and external
dimensions. While Hornsby et al. (2002) focus on a comprehensive internal measure
- 143 -
of corporate entrepreneurship considering management support, organisational
structure, rewards, time and resource availability; Khandwalla (1977) and Zahra
(1991) also include external dimensions such as a competitor orientation and
environmental changes. All three studies demonstrate acceptable reliability values for
their scales, which provide a basis for their use in further research.
Brown, Davidsson and Wiklund (2001:954) developed a measurement instrument
based on Stevenson’s (1983 in Brown et al., 2001:952) conceptual model of
entrepreneurship as opportunity-based firm behaviour.
Brown et al. (2001:955) operationalised Stevenson’s conceptualisation of
entrepreneurial management to assess a firm’s degree of entrepreneurship.
Stevenson’s entrepreneurial management construct consisted of eight dimensions:
strategic orientation, commitment to opportunity, commitment of resources, control of
resources, management structure, reward philosophy, growth orientation and
entrepreneurial culture (Brown et al., 2001:955-956).
The developed scale consisted of a total of 20 items measuring the different
dimensions. Factor analysis showed six distinct factors, which suggests discriminant
validity. Strategic orientation and commitment to opportunity formed one factor and
commitment of resources and control of resources another factor. The remaining
constructs formed one factor each. Reliability measures expressed in Cronbach’s
alpha showed satisfactory values for strategic orientation (alpha = 0.82),
management structure (alpha = 0.78) and growth orientation (alpha = 0.71).
Cronbach’s alpha was lower for entrepreneurial culture (alpha = 0.68), resource
orientation (alpha = 0.58) and reward philosophy (alpha = 0.58) (Brown et al.,
2001:957-959, 963).
Convergent validity was established through a comparison of the entrepreneurial
management scale with Covin and Slevin’s (1989) entrepreneurial orientation scale.
Correlation between the two indices revealed a moderately high degree of
correspondence, which demonstrates that the two measures are related, but only
partly overlapping (Brown et al., 2001:961). Factor analysis with both scales resulted
- 144 -
in nine distinct factors, which represent the six factors of the entrepreneurial
management scale and three factors from the entrepreneurial orientation scale. This
result shows that the two scales are distinct. This was further supported by
correlation analysis of the factor indices, which indicated rather low correlations
between them (Brown et al., 2001:961).
Although the research focus of the two studies by Brown et al. (2002) and Hornsby et
al. (2002) was somewhat different, the concepts that were used for measurement are
comparable. Both approaches include measures of resource availability,
organisational structure, rewards and management support. Furthermore, both
studies demonstrate reliability and validity of each concept. Brown et al. (2001)
further include aspects of strategic orientation, growth orientation and culture. These
additional aspects provide an even more holistic view of the corporate
entrepreneurship construct.
4.2.3 Measuring instruments for entrepreneurial capital
Various studies use three dimensions of entrepreneurial capital, namely financial,
social and human capital (Audretsch & Keilbach, 2004:419; Firkin, 2001:2) for
measuring.
Entrepreneurial capital has been associated with performance and competitive
advantage due to the inimitability of human and social capital (Audretsch & Keilbach,
2004:419; Hatch & Dyer, 2004:1155; Hitt et al., 2001:13).
Financial capital considers financial assets of any form that are directly convertible
into money (Firkin, 2001:2). Audretsch and Keilbach (2004:423) measured financial
capital using inventory and past investments within the manufacturing sector.
Unger et al. (2011:1,6) conducted a meta-analysis on human capital measures that
have been applied in studies over a 38-year period.
- 145 -
Studies were rather diverse in their conceptualisation of human capital, which makes
it difficult to assess what kind of human capital should be considered (Unger et al.,
2011:2).
Quantitative studies that were included in the meta-analysis were grouped in one of
two human capital aspects. First, human capital was considered as an investment
which includes aspects such as education, start-up experience, industry-specific
experience, management experience and work experience. Second, human capital
was considered as an outcome, which summarises effects of human capital
investments such as entrepreneurial skills, competencies and knowledge
(Davidsson & Honig, 2003:306; Unger et al., 2011:3,9).
Unger et al. (2011:10-11) note that not all studies included in the meta-analysis
reported reliability on their measurement. However, the average Pearson Product-
Moment Correlation for studies that reported reliability was 0.77, which indicates high
reliability of the measures.
Hitt and Ireland (2002:4) defined human capital as including aspect such as
education, experience, knowledge and skills. This definition is very similar to that of
Rauch et al. (2005:683,688), who defined human capital as consisting of education,
experience and skills that help to get the work tasks done. The construct was
conceptualised as an index for which the individual dimensions were defined to be
causal. As will be outlined later in this chapter, conventional reliability and validity
assessments cannot be used with causal indicators. Procedures to ensure reliability
and validity of measures were conducted.
Overall, the studies presented by Unger et al. (2011) and Rauch et al. (2005) indicate
that human capital can be measured reliably using the aspects of education,
experience, knowledge and skills.
Davidsson and Honig (2003:307) state that social capital refers to people’s ability to
take advantage of their social structures, networks or memberships. Furthermore,
social capital is considered to be multidimensional and occurs on an individual and
organisational level.
- 146 -
Social capital has been operationalised using the number of organisations,
associations, communities or alumni organisations an individual belongs to
(Davidsson & Honig, 2003:309,314; De Carolis et al., 2009:535).
De Carolis et al. (2009:535) operationalised social capital using two dimensions:
social networks and relational capital. The concept of social networks was measured
by the number of networks a person belonged to. The second dimension consisted of
three items measuring the extent of involvement in these organisations. Reliability of
social capital was acceptable, with a Cronbach’s alpha of 0.76.
Baron and Markman (2000:107) suggest that social capital should also consider
social skills. They argued that a person’s access to networks depends on the ability
to interact effectively with others.
Baron and Markman (2003:49) measured social skills by various items suggested by
the social skills inventory (SSI) by Riggio (1986:652), and further developed new
items. The resulting factor structure considered social perception, social adaptability,
and expressiveness, which yielded acceptable Cronbach’s alpha values of 0.83, 0.67
and 0.74 respectively.
The social skills inventory (SSI) consists of 105 items which measure seven
dimensions of social skills: emotional expressivity, emotional sensitivity, emotional
control, social expressivity, social sensitivity, social control and social manipulation.
The dimensions showed good reliability values; Cronbach’s alpha was between 0.75
and 0.88 (Riggio, 1986:653).
Overall the studies on social capital presented indicate that the construct can reliably
be measured in its quantitative and qualitative aspects. The quantitative part includes
aspects such as the number of networks or the amount of time spent conducting
networking activities. The qualitative aspects of the construct consider social skills,
which include social perception, adaptability or expressiveness (Baron & Markman,
2003; Davidsson & Honig, 2003; De Carolis et al., 2009; Riggio, 1986).
- 147 -
4.2.4 Measuring instruments for market orientation
The two main scales for market orientation have been developed by Narver and
Slater (1990) and Kohli and Jaworski (1990). Desphandé et al. (1993) developed a
scale to assess a firm’s customer orientation.
The three scales have been further developed and streamlined by Deshpandé and
Farley (1998) and Matsuno et al. (2005).
The following paragraphs outline the properties of the different market orientation
scales.
Narver and Slater (1990:22) developed a measurement instrument for market
orientation as an organisational culture.
Narver and Slater (1990:22) defined market orientation as a unidimensional construct
consisting of three behavioural aspects: customer orientation, competitor orientation
and interfunctional coordination, and two decision criteria: long-term focus and profit
objective of the business (Narver & Slater, 1990:22).
Items for each of the five concepts of market orientation were developed, and content
validation was established by using a panel of experts in the strategic marketing field
(Narver & Slater, 1990:23). Customer orientation was measured using six items,
competitor orientation included four items, and interfunctional coordination consisted
of five items. The two decision criteria were measured using three items each
(Narver & Slater, 1990:24).
Internal consistency was assessed using Cronbach’s alpha and item-to-total
correlations. The Cronbach’s alpha values for customer orientation (alpha = 0.868),
competitor orientation (alpha = 0.727) and interfunctional coordination
(alpha = 0.735) were satisfactory. However, values for long-term focus
(alpha = 0.408) and profit emphasis (alpha = 0.004) were below the recommended
value of 0.7 (Nunnally, 1978 in Santos, 1999:2). Consequently the two decision
- 148 -
criteria were not further analysed. However, inter-rater reliability was assessed and
showed satisfactory results (Narver & Slater, 1990:24).
Construct validity was assessed by convergent, discriminant and concurrent validity.
Convergent validity between the three concepts of market orientation was assessed
with correlation analysis, which showed satisfactory results. Discriminant validity was
measured with a scale that is considered to be different from market orientation. The
results provided support for discriminant validity (Narver & Slater, 1990:25).
To assess concurrent validity, Narver and Slater (1990:26) correlated the market
orientation construct with two constructs that had been validated before. The
correlations showed satisfactory results, providing support for concurrent validity
(Narver & Slater, 1990:26).
Jaworski and Kohli (1993:53) constructed their scale of market orientation using
three different concepts: intelligence generation, intelligence dissemination and
responsiveness to information. A total of 32 items were developed to assess the
concepts. 10 items relate to intelligence generation, eight to intelligence
dissemination and 14 to responsiveness to information.
Reliability measures in the form of Cronbach’s alpha for the dimensions showed
satisfactory results. The following Cronbach’s alpha values were identified:
intelligence generation 0.71, intelligence dissemination 0.82, responsiveness to
information 0.82 (Jaworski & Kohli, 1993:60,65).
A mean score for the overall market orientation construct was calculated by adding
the corresponding item scores from all three concepts. Correlation between the
overall score and each of the three concepts, as well as correlation between the
three concepts, showed satisfactory results (Jaworski & Kohli, 1993:60).
In another study Kohli et al. (1993:467) developed the MARKOR scale, which is a
further refinement of the original scale by Jaworski & Kohli (1993).
- 149 -
It hypothesised that the market orientation construct represents one general factor
consisting of three correlated factors. Based on error variance estimates and analysis
of cross-loadings of items, the original scale of 32 items was reduced to 20 items,
which represented a better model fit (Kohli et al., 1993:470).
In another step a multi informant sample was drawn to run a replication analysis to
determine the appropriate factor structure. It was found that several of the models
that included a general factor of market orientation and three component factors
lacked discriminant validity between intelligence dissemination and responsiveness
to information (Kohli et al., 1993:470-471).
Further confirmatory factor analysis was conducted to assess the validity of the
market orientation construct. Overall moderate validity was found (Kohli et al.,
1993:473).
Deshpandé et al. (1993:24,27) developed a measure for customer orientation. The
dimensions of customer orientation related to the conceptual definition by Kohli and
Jaworski (1990) and Narver and Slater (1990). The construct consisted of nine items
(Deshpandé et al., 1993:29,33-34).
Reliability was assessed using Cronbach’s alpha, which was satisfactory
(alpha = 0.69). Internal validity was determined using item-to-total correlations
(Deshpandé et al., 1993:29).
Deshpandé and Farley (1998:213,216) directly compared the three scales developed
by Narver and Slater (1990), Kohli et al. (1993) and Deshpandé et al. (1993), based
on reliability and validity analysis. Moreover, a synthesis of the three scales for
market orientation into the MORTN scale was developed.
Acceptable Cronbach’s alpha levels were found for the Narver and Slater scale
(1990) (alpha = 0.90) and the Deshpandé et al. scale (1993) (alpha = 0.72) and
somewhat lower reliability levels for the Kohli et al. scale (1993) (alpha = 0.51)
(Deshpandé & Farley, 1998:216).
- 150 -
Construct validity was calculated using three items from an unrelated scale. Results
showed strong discriminant validity (Deshpandé & Farley, 1998:218). External
validity was assessed using a constant-sum scale which showed satisfactory results
for all three scales (Deshpandé & Farley, 1998:217). Predictive validity considering
performance indicators for all three scales was determined, showing satisfactory
results (Deshpandé & Farley, 1998:218).
High correlations between the three scales, as well as a high degree of intra-
company reliability, could be shown. A comparison of inter-rater reliability allows the
conclusion that the three scales can be used interchangeably in practice
(Deshpandé & Farley, 1998:218-219).
A cross-national comparison of the three scales showed strong reliability in European
and US studies. The scale by Deshpandé et al. (1993) has the broadest international
exposure, with applications of the scale in India, China, Japan, Germany and
England (Deshpandé & Farley, 1998:219-220).
Cross national assessment of validity did not show any significant differences
between countries or industries, lending support to the conclusion that all three
scales are valid across different nations and industries (Deshpandé & Farley,
1998:221-222).
A synthesis of the three scales was developed in order to account for redundancies
by using all three scales as well as achieving a smaller number of items to make the
market orientation scale practical for use in larger studies (Deshpandé & Farley,
1998:222).
Factor analysis with 44 items from all three scales was performed resulting in one
factor that explained more than 40% of variance. This factor included 10 items from
all three scales and was termed the MORTN scale. The MORTN scale showed high
reliability (alpha = 0.88) and predictive validity (Deshpandé & Farley, 1998:222-223).
Matsuno et al. (2005:3-4) developed the EMO (extended market orientation) scale,
which extends the scales by Narver and Slater (1990) and Kohli et al. (1993). A
- 151 -
comparison between the EMO scale and the scales by Narver and Slater (1990) and
Kohli et al. (1993) was conducted.
The EMO scale incorporates a broader scope of stakeholders and market factors. It
consists of the concepts of intelligence generation, dissemination activities and
responses to market players and has a total of 22 items. Information generation
consisted of eight items, information dissemination included six items and
responsiveness to information was measured with eight items. Cronbach’s alpha for
each concept and the overall market orientation measure (alpha = 0.85) showed
satisfactory results: information generation (alpha = 0.65), information dissemination
(alpha = 0.75), responsiveness to information (alpha = 0.81). Further, convergent
validity was established (Matsuno et al., 2005:4-6).
Unidimensionality for all three scales was assessed. A second-order confirmatory
factor analysis produced satisfactory fit indices for the EMO and the Narver and
Slater (1990) scale. For the Kohli et al. (1993) scale unidimensionality could not be
assessed and it was subsequently removed from further analysis (Matsuno et al.,
2005:5).
On the level of the second-order factor structure the two remaining scales were
assessed based on their fit statistics. Both scales achieved good fit statistics
(Matsuno et al., 2005:5).
Predictive validity was determined for the Narver and Slater (1990) scale and the
EMO scale using performance measures in structural equation modelling. Both
scales were positively related to performance indicators indicating predictive validity.
However, the scale by Narver and Slater (1990) was considered to be more efficient
in prediction as it considers fewer items (Matsuno et al., 2005:5-6).
Overall, Matsuno et al. (2005:6) note that no single scale was found absolutely
satisfactory.
The presented studies on measuring instruments for market orientation have evolved
around the concepts developed by Narver and Slater (1990), Kohli and Jaworski
- 152 -
(1990) and Jaworski and Kohli (1993). Various studies extended and compared the
original scales (Deshpandé et al., 1993, Deshpandé & Farley, 1998; Kohli et al.,
1993; Matsuno et al., 2005). The studies showed consistently acceptable reliability
and validity of the scales, even in an international setting (Deshpandé et al., 1993).
However, it appears that no scale is completely satisfactory in capturing the construct
of market orientation (Matsuno et al., 2005:6). It needs to be noted that market
orientation has been measured as a reflective construct that is considered to be
either unidimensional (Deshpandé & Farley, 1998; Narver & Slater, 1990) or
multidimensional (Jaworski & Kohli, 1993; Kohli et al., 1993; Matsuno et al., 2005).
4.2.5. Measuring instruments for firm performance and relative competitive
strength
Financial performance can be measured using objective measures or subjective
measures. Objective measures relate to actual percentage figures of sales growth,
turnover or profitability (Dawes, 1999:65). However, as mentioned in chapter three,
most managers are unwilling to disclose firm performance indicators (Moorman &
Rust, 1999:187). Moreover, in some cases the collection of objective financial data
may not be viable, as the data may only be available on an aggregated level which is
not appropriate for the level of analysis (Wall, Michie, Patterson, Wood, Shehan,
Clegg & West, 2004:96).
Various studies measuring market orientation have previously used subjective
measures of performance (Dawes, 1999; Dess & Robinson, 1984; Moorman & Rust,
1999).
Measures for subjective performance relate to questions such as “Please rate the
overall financial results of your firm”, and “Please rate the return on investment or
return on assets of your firm”. Answers to these questions are given on a scale with
anchor labels such as “very good”, “very poor” (Dawes, 1999:65,69-70).
A comparison of studies that use objective and subjective performance measures
shows that these measures are strongly correlated, which demonstrates convergent
- 153 -
validity (Dawes, 1999:68-70; Dess & Robinson, 1984:269; Moorman & Rust,
1999:187).
Wall et al. (2004:95) analysed the validity of subjective measures. The subjective
measures asked for the company’s performance in comparison to the main
competitor. Objective measures included financial data from audited records. The
study showed convergent and discriminant validity (Wall et al., 2004:101,104,111).
Considering the results of the presented studies, subjective measures of financial
firm performance provide a reliable measure if objective data cannot be obtained.
Some researchers (Barney, 1991:99) argue that competitive advantage derives from
the firm’s resources, which must be valuable, rare, inimitable and sustainable. Other
researchers (Porter, 1980 in Cockburn et al., 2000:1126) argue that competitive
advantage derives from the firm’s microeconomic environment (Cockburn et al.,
2000:1126).
Cockburn et al. (2000:1128) state that competencies may lead to competitive
advantage. However, one also needs to understand where the competencies come
from.
In order to achieve competitive advantage it is necessary to have the required
resources but at the same time have strategies to transform these resources into
capabilities (Chandler & Hanks, 1994:335).
Generic strategies of competitive advantage are described as cost leadership,
differentiation and focus (Porter, 1998:xxii).
Chandler and Hanks (1994:338) measured competitive strategies using three
dimensions: innovation, quality and cost leadership. All three dimensions were
measured with multiple items.
Innovation was measured by three items: being the first to have new products
available, stressing new product development and engaging in innovative marketing
- 154 -
techniques (Chandler & Hanks, 1994:338). The second dimension, high quality, was
measured by five items: emphasis on quality control, meeting customer
requirements, emphasis on firm’s superior customer service, producing only the
highest quality products and setting customer needs first (Chandler & Hanks,
1994:338). Cost leadership was the third dimension, consisting of three items:
emphasising cost reduction in business operations, emphasising improvement in
employee productivity and operations efficiency, and lower production cost due to
process innovation (Chandler & Hanks, 1994:338).
For all three dimensions Cronbach’s alpha was assessed and showed acceptable
values: innovation (alpha = 0.70), quality (alpha = 0.78) and cost-leadership
(alpha = 0.73). Discriminant validity between the three dimensions was also
established (Chandler & Hanks, 1994:338-339).
Zhou, Brown and Dev (2009:1065) measured competitive advantage using
differentiation advantage consisting of two types: namely market differentiation and
innovation differentiation. Items for market and innovation differentiation were
modified from Chandler and Hanks (1994). Exploratory and confirmatory factor
analysis were performed to assess reliability and convergent and discriminant
validity. The analysis showed acceptable values. Composite reliability for market
differentiation was 0.73 and for innovation differentiation 0.66 (Zhou et al.,
2009:1067-1068).
Burke (1984:347) developed a measure for relative competitive strength which
considered a business unit’s share of the market. This measure is considered to
reflect the business unit’s position within the market compared with that of major
competitors.
Relative competitive strength was measured by multiple items. Items compared the
business unit with its major competitor in five dimensions: product changes, price
changes, service improvements, technological innovation and marketing methods.
Reliability was assessed by Cronbach’s alpha, which was completely satisfactory
(alpha = 0.94). Discriminant validity between relative competitive strength and other
constructs was established (Burke, 1984:351-353).
- 155 -
Competitive strength is considered to have explanatory power over performance. It
captures how well the firm systematically engages with its environment and how well
it can distinguish itself from other firms (Augusto & Coelho, 2009:96).
Augusto and Coelho (2009:96-98,101) state that competitive strength captures
aspects of how well a firm can anticipate and shape the market in which it operates.
Competitive strength was measured by five items relating to those of the competition:
the organisation’s prices, quality of products, capacity to compete, diversity of
product and its tendency to be ahead of competitors. The items were derived from
Burke (1984). Composite reliability, which can be compared to Cronbach’s alpha,
was acceptable, with a value of 0.81 which exceeds the reference value of 0.7.
Convergent and discriminant validity were established (Augusto & Coelho,
2009:100).
The presented studies on competitive advantage measure the construct using
different aspects of market-related items and firm-internal items. The market-related
items refer to market share and activities to compete with competitors. Firm-internal
factors consider the capacity to innovate, and quality and cost aspects. In
combination these items provide a good measure of the construct, as has been
demonstrated by the acceptable reliability and validity measures (Augusto & Coelho,
2009; Burke, 1984; Chandler & Hanks, 1994; Zhou et al., 2009).
4.3 LITERATURE REVIEW ON STATISTICAL MODELLING
In statistical modelling, causal modelling is considered to be the most prominent
approach for theory development. This framework considers cause and effect
relationships between constructs (Jaccard & Jacoby, 2010:137-138; Shmueli,
2010:289). Structural equation modelling (SEM) is used to quantitatively assess
cause-effect relationships between variables of interest (Pearl, 2007:135).
In statistical modelling a careful distinction between causal explanation and empirical
prediction needs to be made (Shmueli, 2010:289). The purpose of exploratory
modelling is causal explanation, which tries to match the statistical model with the
data to draw inferences (Shmueli, 2010:290,293). Empirical prediction is conducted
- 156 -
in predictive modelling. The focus of predictive modelling is on the individual
constructs in the model. The statistical model is used to predict new values of the
dependent variable (Shmueli, 2010:290,293). The perspectives of causation versus
prediction are also reflected in the two approaches that can be applied to SEM,
namely covariance-based structural equation modelling and partial-least squares
modelling. The two approaches will be discussed in more detail in chapter five.
The general model of SEM considers two types of variables: latent variables or latent
constructs, which are variables that cannot be directly observed or measured. These
variables have also been termed unobserved or unmeasured variables. Latent
variables need to be inferred from a set of observed variables. Observed variables
are also called measured, manifest, or indicator variables, items or proxies. These
variables represent a set of variables that are used to define the latent variable or
construct (Bollen, 1989:11,16; Diamantopoulos et al., 2008:1204; Schumacker &
Lomax, 2010:3).
SEM consists of two parts. The first is a measurement model that specifies the
relationships between the latent variables and their measures. The second is a
structural model which specifies the causal relationships between the latent variables
(Anderson & Gerbing, 1988:411; Bagozzi & Baumgartner, 1994:387; Bollen, 1989:11;
Burke Jarvis et al., 2003:199; Diamantopoulos et al., 2008:1204; Edwards & Bagozzi,
2000:155; Law et al., 1998:741).
The following paragraphs outline both parts and describe the various steps to specify
the measurement model and the structural model.
4.3.1 Measurement model
Multidimensional constructs are often used in research to assess the overall latent
variable. A multidimensional construct is a construct involving more than one
dimension which is treated as a single theoretical concept. A dimension is a
manifestation of the construct (Diamantopoulos et al., 2008:1205; Edwards,
2001:144; Law et al., 1998:741; Law & Wong, 1999:144; MacKenzie et al.,
2005:711).
- 157 -
It is argued that multidimensional constructs are useful as they present a holistic view
of complex phenomena and at the same time account for precision in measuring the
dimensions (Edwards, 2001:145). However, it is important that the relationships
between the construct and its dimensions are well defined (Law et al., 1998:741).
There are various ways in which a multidimensional construct can relate to its
dimensions and the dimensions to the indicators. When dealing with
multidimensional constructs it is necessary to distinguish between different levels of
analysis. The first level (first-order) is relating the observed variables to their
dimensions. The second level (second-order) relates the dimensions to the latent
constructs. For each level either a formative or a reflective specification is possible
(Diamantopoulos et al., 2008:1205-1206)
The following figure presents the two types of relationships.
FIGURE 4.1: Reflective and formative relationships
Reflective model Formative model
Dimension 1
Latent construct
First-levelrelationships
Dimension 2 Dimension 3 Dimension 1
Latent construct
Dimension 2 Dimension 3
Second-levelrelationships
indicators indicators indicators indicators indicators indicators
Measurement error(Ɛ1)
Measurement error(Ɛ2)
Measurement error(Ɛ3)
Error term(ζ)
Error term(ζ)
Error term(ζ)
Error term(ζ)
Source : adapted from MacKenzie et al. (2005:711,714)
- 158 -
Burke Jarvis et al. (2003:204) observe that it is possible for single multidimensional
constructs to have different measurement models, one relating its measures to its
first-order constructs and another one relating its dimensions to the second-order
factor. Therefore it is possible to have mixed models such as first-order reflective and
second-order formative or vice versa.
The characteristics of the reflective and formative models can be described from four
aspects, such as causality, intercorrelation, error term, antecedents and
consequences. The following descriptions apply to both, first-level and second-level
relationships, but only second-level relationships will be explicitly described.
First, in the reflective model the causality flows from the construct to the dimensions.
Therefore the structural paths on the path diagram (Figure 4.1) point from the
construct towards the dimensions (Bollen & Lennox, 1991:306; Burke Jarvis et al.,
2003:203; Diamantopoulos et al., 2008:1205; Law & Wong, 1999:144-145;
MacCallum & Browne, 1993:533; MacKenzie et al., 2005:711,713).
In the formative model the dimensions cause the construct, they make the construct
appear. The paths in the path diagram (Figure 4.1) point from the dimensions to the
construct (Edwards, 2001:147; Law et al., 1998:745; Law & Wong, 1999:146;
MacCallum & Browne, 1993:533).
Second, the dimensions in the reflective model need to be positively correlated as
they represent the same underlying construct and share a common theme (Bollen &
Lennox, 1991:307; Burke Jarvis et al., 2003:203; Diamantopoulos et al., 2008:1205;
Law & Wong, 1999:144-145; MacKenzie et al., 2005:711).
In formative models there are no specific expectations about intercorrelations
between dimensions (Bollen & Lennox, 1991:308; Diamantopoulos et al.,
2008:1204). Formative dimensions of the same construct can have positive or
negative correlations or no correlation. As the dimensions capture distinct aspects of
the latent construct, they are not interchangeable. Omitting one dimension changes
the whole construct (Bollen & Lennox, 1991:308; Diamantopoulos et al., 2008:1204).
- 159 -
Therefore using instruments like factor analysis to examine the correlations between
the dimensions could have a serious negative impact, as one could overlook valid
dimensions that determine the construct. Moreover, high correlations between the
dimensions make it difficult to separate the distinct influence of each dimension on
the latent construct, which is described as a multicollinearity problem (Bollen &
Lennox, 1991:307).
Third, in the reflective model the measurement error is determined on the level of
observed variables (Bollen & Lennox, 1991:306; Diamantopoulos & Winkelhofer,
2001:271; MacKenzie et al., 2005:711).
For formative models an error term is captured on the construct level which impacts
on the latent variable but is uncorrelated with the observed measures. The error term
cannot be considered as a measurement error. It is rather a disturbance term that
comprises all remaining causes of the construct that are not represented by the
indicators (Diamantopoulos, 2006:9-10; Edwards, 2001:155).
Fourth, antecedents and consequences of the measures need to be considered. In
the reflective model indicators reflect the underlying construct and should therefore
have the same antecedents and consequences (Burke Jarvis et al., 2003:203;
MacKenzie et al., 2005:713). In formative models indicators do not necessarily
capture the same aspects of the construct and therefore they cannot be expected to
have the same antecedents and consequences (Burke Jarvis et al., 2003:203;
MacKenzie et al., 2005:713).
4.3.1.1 Scale development
Rossiter (2002:306,308) describes a six-stage process of generating and selecting
items to measure a construct. The process considers reflective and formative cases
in scale development. The steps for scale development are outlined in the following
paragraphs.
First, the construct must be defined in terms of the object, the components and the
attributes, and the rater entity. The construct refers to a phenomenon of theoretical
- 160 -
interest that is described in terms of the object, including its constituents or
components (reflective or formative dimensions), the attributes (reflective or
formative indicators) and the rater (person who judges) (Rossiter, 2002:308,310).
The second step in scale development is to classify the objects. Objects can either
be concrete singular (unidimensional), abstract collective (reflective
multidimensional) or abstract formed (formative multidimensional). The latter two
classifications require an index when it comes to enumeration and reporting
(Rossiter, 2002:313).
Third, attributes can be classified into concrete (singular), (abstract) formed and
(abstract) eliciting (Rossiter, 2002:313).
Concrete attributes have unanimous agreement between different raters and refer to
only one characteristic. For these attributes a single-item measure is sufficient and
valid. This means that the description of that item and the response categories must
be clear (Rossiter, 2002:313-314).
Components adding up to the overall meaning of the attribute are called formed
attributes. The response to the components causes the attribute to appear (Rossiter,
2002:314).
Researchers have different opinions about the number of attributes. Diamantopoulos
and Winkelhofer (2001:271) state that formed attributes require a census of
indicators, whereas Rossiter (2002:314) argues that the formed attribute needs to
include main components; otherwise one searches for low-incidence components.
An abstract-eliciting attribute describes attributes that consider traits or a disposition.
The items are manifestations of the trait or disposition. Furthermore, raters’ answers
on which characteristics represent the attribute would differ only slightly. Eliciting
attributes should be written as a set of distinct activities. Items are interchangeable
and a reasonable sample of items, up to five items, is considered to be sufficient
(Rossiter, 2002:316-318).
- 161 -
The next step in scale development is rater identification. The scale score reliability
will differ depending on the rater entity, which can be an individual, experts or a
group. First, individual raters are persons providing self-reports on attributes. Expert
raters, conduct a content analysis and thus reliability of the attributes. Group raters
are usually a sample of consumers, managers, salespersons etc. The object that
these raters assess is often a company or a product (Rossiter, 2002:318-319).
The fifth step in scale formation puts together the object items with their attribute
items to form the scale. The scale items need to be easily understood, and this
needs to be tested in pre-tests. For first-order eliciting attributes, coefficient beta,
which is a test to assess unidimensionality, should be computed. Values of at least
0.5 are needed to infer that there is a general factor accounting for 50% of the item
variance. For second-order eliciting attributes, a confirmatory factor analysis can be
applied (Rossiter, 2002:320-322). The response answer format should consider
questions that do not imply any intensity. The answer categories should be
developed considering minimum to maximum intensity (Rossiter, 2002:323).
The last step in scale development is the enumeration process. This process
describes procedures to derive a total score from scale items. As the construct can
consist of different object and attribute types, the procedures will vary. An index can
be described as a sum of item scores. A profile rule can also be established where
each component must exceed a minimum level in order to be included in the index. A
multiplicative rule can be applied in cases where a theory regarding the construct’s
algebraic relations is available. In all other cases a linear relationship between the
construct and its dimensions should be assumed. Items can also be weighted before
the index is computed. However, it is necessary to have a conceptual definition for it,
as empirical weighting is not appropriate. Furthermore, items for indexes cannot be
deleted, as they form the scale. Eliciting attributes are, however, interchangeable
items (Law et al., 1998:751; Rossiter, 2002:325).
- 162 -
4.3.1.2 Reliability and validity assessments of the measurement model
Bollen (1989:194) states that the problem with common reliability and validity tests is
that they consider only observed measures but do not account for the latent variable
and hence measurement error.
In order to determine reliability and validity of the measurement model it is necessary
to assess the indicators themselves, as well as their relation to their latent variables.
The procedures vary depending on the type of indicators, reflective versus formative
(Jahn, 2007:21).
In the reflective case, a first step is to analyse factor loadings, which should achieve
a minimum value of 0.7 (Henseler, Ringle & Sinkovics, 2009:299; Jahn, 2007:21).
However, Chin (1998:325) notes that loadings with 0.5 and 0.6 can also be
considered if research development is in the early stages.
In a second step internal consistency of indicators is determined with composite
reliability. Composite reliability is more accurate than Cronbach’s alpha, which is
sensitive to the number of indicators (Chin, 1998:320; Jahn, 2007:21).
A third step combines reliability and validity assessments in the measurement model.
The average variance extracted (AVE) determines the amount of variance that is
captured by the construct in relation to the amount of variance due to measurement
error (Fornell & Larcker, 1981:45). Values lower than 0.5 indicate that the variance
due to measurement error is larger than the variance captured by the construct. This
means that validity of the indicators and the construct is questionable (Fornell &
Larcker, 1981:46).
In the case of formative indicators reliability in the form of internal consistency cannot
be applied as indicators can have positive, negative or zero correlation (Burke Jarvis
et al., 2003:202; Diamantopoulos et al., 2008:1215).
Indicator validity can be assessed in various ways. First, the loadings which reflect
the impact of the formative indicator on the latent construct need to be significant.
- 163 -
Items with non-significant loadings should be eliminated as they do not represent
valid indicators of the construct (Diamantopoulos et al., 2008:1215).
Diamantopoulos and Winkelhofer (2001:272) suggest validity assessment by using
an overall measure that summarises the essence of the construct. A high relationship
between the formative indicator and the overall item signifies indicator validity.
Assessing validity on the construct level focuses on nomological and/or criterion-
related validity. To assess nomological validity, Diamantopoulos and Winkelhofer
(2001:273) suggest linking the latent construct with other related constructs such as
antecedents and/or consequences. In order to assess validity three steps need to be
considered. First, it is necessary to gather additional information on the related
construct. Second, the related construct needs to be reflective, and third, a
theoretical relationship between the latent construct and the related construct needs
to be generated.
Diamantopoulos and Siguaw (2006:271-272) provide an approach to assessing
criterion validity. An external construct which is positively related to the focal
construct is developed. Regression analysis is performed on all indicators of the focal
construct (reflective and formative) and the external construct to assess the
magnitude and significance of the relationships.
Diamantopoulos and Siguaw (2006:267,275) note that the initial item pool for both
types of measurement, reflective or formative, is the same. Hence, items that have
been used in reflective models can also be applied to formative models. However,
after identification, reliability and validity assessments have been conducted, one
cannot expect to have the same item pool for reflective and formative models.
The previous paragraphs have outlined reliability and validity assessments of
measurement models. It has been noted that different approaches need to be
followed when dealing with reflective and formative models. The reflective
assessment basically follows traditional scale evaluation procedures, which allow for
determining reliability and validity. For formative models reliability cannot be
determined. However, validity becomes even more important, and this was described
- 164 -
by the assessment of validity measures such as indicator validity, nomological
validity and criterion validity. Further, formative models need to be identified, which
requires additional measures (Diamantopoulos & Winkelhofer, 2001; Diamantopoulos
& Siguaw, 2006; Diamantopoulos et al., 2008).
4.3.1.3 Model misspecification and its impact
Although reflective and formative measures were initially developed around the same
time, the 1960s and 1970s, in today’s social sciences reflective measurements are
commonly used, whereas formative measurements are rarely used in research
endeavours (Diamantopoulos, 2006:7). A reason for this may be the convenience of
analysing models under the reflective view. A number of programmes are available
for the analysis of covariance based structural equation modelling (Law & Wong,
1999:156).
However, researchers argue that technical convenience should not guide the
research and the adoption of the type of measurement model as a misspecification
can have serious effects (Law & Wong, 1999:159).
Model misspecification influences the estimates of the measurement and structural
model parameters, which affect the conclusion about the theoretical relationships
among the constructs (Burke Jarvis et al., 2003:207,209). Burke Jarvis et al.
(2003:212) conducted a Monte Carlo simulation manipulating the measurement
model and the structural model. It was found that goodness-of-fit indices are not able
to detect model misspecification, as indices produce satisfactory results for the
incorrectly as well as the correctly specified models. A model could show satisfactory
fit indices even though the structural parameters are biased, which would result in
misleading inferences. Furthermore, paths in the structural model coming from a
misspecified construct could lead to type I errors. Paths leading into a construct with
a misspecified model could lead to type II errors.
MacKenzie (2003:324) notes that measurement model misspecification can
undermine construct validity. First, the relationship between the measures and the
construct are misrepresented. Second, if a formative indicator were treated as a
- 165 -
reflective indicator scale purification methods, such as alpha coefficients, could lead
the researcher to drop items even though they represent valid measures of the
construct.
Burke Jarvis et al. (2003:206) reviewed the measurement model specifications of
four top marketing journals regarding construct definition over a 24-year period. Their
results indicated that out of 1192 analysed constructs, 839 (70%) were correctly
specified, and 353 (30%) were incorrectly modelled. The majority of constructs (810)
were reflective constructs which were correctly specified.
Podsakoff et al. (2003:649-650) found that out of 138 analysed leadership constructs,
65 (47%) were incorrectly specified, with reflective indicators rather than the correct
specification of formative indicators.
Burke Jarvis et al. (2003:208) present an overview of constructs commonly used in
marketing literature which should be specified in a formative way. Among those
constructs is market orientation, which should be specified as a second-order
formative construct involving intelligence generation, dissemination and
responsiveness to information. As outlined in section 4.2.4, the market orientation
construct has previously been measured as a unidimensional (Narver & Slater, 1990)
or reflective multidimensional construct (Jaworski & Kohli, 1993; Kohli et al., 1993;
Matsuno et al., 2005).
Cadogan, Souchon and Procter (2008:1263) developed a formative,
multidimensional model of market-oriented behaviours using the elements of the
market orientation construct by Kohli and Jaworski (1990). Cadogan et al.
(2008:1272-1274) identified all the variables that form market orientation and
assessed nomological validity for the model. The fit indices for all three concepts of
information generation, dissemination and responsiveness were good. Cadogan et
al. (2008:1274) suggest replicating the study in order to prove stability.
Coltman, Devinney, Midgley and Venaik (2008:1260) also modelled market
orientation as a formative construct. In their empirical study it was found that market
orientation can be modelled in both ways: reflective and formative.
- 166 -
George (2011:12-15) develops a second-order formative model of entrepreneurial
orientation, consisting of the reflective first-order concepts innovation, proactiveness
and risk-taking. The formative model is compared with a reflective model. The results
indicate that when entrepreneurial orientation is constructed as a second-order
reflective construct, the causal paths are inflated, leading to invalid conclusions.
However, in a comparison of fit indices the model misspecification could not be
detected.
The presented studies indicate that model misspecification is a serious problem that
can lead to wrong inferences. Furthermore, it has been found that fit indices are not
always suitable for detecting model misspecification (Burke Jarvis et al., 2003;
George, 2011; MacKenzie, 2003). A selected number of studies were presented that
outline the modelling of market orientation and entrepreneurial orientation as
formative constructs. The studies showed acceptable reliability and validity values
(Cadogan et al., 2008; George, 2011). One study found that market orientation can
be modelled either in a formative or reflective way (Coltman et al., 2008).
4.3.2 Structural model
The structural model focuses on the causal relationships between the latent variables
which are represented as paths (Bollen, 1989:11; Jaccard & Jacoby, 2010:166).
First, the direction of the paths needs to be determined, which is followed by an
estimation of the path strength. Path strength is described by path coefficients (Jahn,
2007:10).
Two types of structural models can be distinguished, which determine the demands
for statistical analysis. First, the recursive model is characterised by uncorrelated
disturbance terms and unidirectional causal effects. The statistical requirements for
this analysis are rather simple. Second, non-recursive models have correlated
disturbance terms and can have feedback loops which require additional
assumptions (Kline, 2011:106). For the purpose of this study a recursive model of
market-driving ability was developed.
- 167 -
Structural models distinguish between exogenous and endogenous variables. An
exogenous variable does not have causal paths pointing at it, whereas endogenous
variables have at least one causal path going into them (Jaccard & Jacoby,
2010:145). The effect of one latent variable on the other can be analysed as direct,
indirect and total effects (Jahn, 2007:10).
In a final step the whole model (structural and measurement) is submitted to testing.
The available procedures and the appropriate fit indices to evaluate model fit will be
outlined in chapter five.
The following section will outline the measurement and structural model for
market-driving ability in corporate entrepreneurship.
4.4 MODEL OF MARKET-DRIVING ABILITY IN CORPORATE
ENTREPRENEURSHIP
In the following the measurement and structural model of market-driving ability in
corporate entrepreneurship are described.
4.4.1 Measurement models
The following section outlines the measurement models used in this study to
determine market-driving ability. The scale and index development follows the steps
described by Rossiter (2002).
The rater identity is the same for all following measurement models. Raters in this
study are members of an organisation operating in the South African healthcare
industry who hold a junior, middle or top management position.
The South African healthcare sector is characterised by a dual system: the public
and private sector. The overall spending of the public sector accounts for 34% of total
health expenditure, while the private sector makes up 66%. The prevalence of
diseases like HIV/Aids and tuberculosis is among the highest in the world (Avert,
n.d.). Furthermore, diseases of a modern society such as hypertension and diabetes
- 168 -
are very evident. The South African government strives to improve the healthcare for
all citizens and plans to introduce a national health insurance system
(SouthAfrica.info, 2009).
The healthcare sector comprises several market players which form a part of this
study, such as the pharmaceutical industry, medical device industry, medical
schemes and pharmaceutical distributors/wholesalers. The healthcare sector
provides growth opportunities as well as challenges due to changes in the regulatory
environment, which makes it an ideal industry to investigate market-driving ability of
organisations.
4.4.1.1 Measurement model for market driving
As outlined in chapter three researchers differ in their opinions on how market driving
can be described. For the purpose of this study market driving is considered to be a
second-order formed abstract object which includes three components: a firm’s
activities regarding market sensing, influencing customer preferences and alliance
formation.
Market driving is considered as formative, as the three dimensions make unique
contributions to the construct. Hence, omitting one would change the construct of
market driving. Furthermore, a change in one of the three components would be
expected to change the overall construct of market driving. Finally, the three
components of market sensing, influencing customer preferences and alliance
formation do not share a common theme (MacKenzie, 2003:325).
Following Rossiter’s (2002:313) suggestion, an index was generated for market
driving.
The next step in scale formation was to classify the dimensions market sensing,
influencing customer preferences and alliance formation. The items of these
dimensions are first-order relationships and represent the specific manifestation of
the dimension; hence they are described as eliciting attributes.
- 169 -
The items of the respective dimensions are indicative rather than formative. Causality
flows from the dimensions, such as market sensing, to the items which represent the
dimension.
As suggested by Rossiter (2002:317) each attribute should include three to five items
to assess unidimensionality. In accordance with the discussed literature (Bollen &
Lennox, 1991:306; Diamantopoulos & Winkelhofer, 2001:271; MacKenzie et al.,
2005:711), for each observed reflective item a measurement error term is added and
a disturbance term is added for the formative construct of market driving.
The full questionnaire is provided in Annexure A.
Alliance formation was measured by five self-constructed items (questions 54-58) for
which ideas were taken from Kale et al. (2000), Baron and Markman (2000) and
Gulati (1999).
Market sensing was measured by five items (questions 61-65) which were adapted
from the scanning intensity scale by Barringer and Bluedorn (1999) and the scanning
items used by Miller and Friesen (1982).
Influencing customer preferences was measured by four self-constructed items
(questions 50-53), taking ideas from Narver et al. (2004), Jaworski et al. (2000) and
Kumar et al. (2000).
The last step in scale formation has been described as the enumeration process
(Rossiter, 2002:325). A summed index was formed for market driving which was
derived from the sum of item scores. Further, a linear relationship between
market-driving and its components was assumed.
Figure 4.2 summarises the measurement model for market driving.
- 170 -
FIGURE 4.2: Measurement model for market driving
ALL
SENS MD
Q 61
Q 62
Q 63
MD Market-driving ALL Alliance formation SENS Market-sensingCUST Customer preferences
ε Measurement error (reflective)ζ Disturbance term (formative)
CUST
Q 50
Q 51
Q 52
Q 64
Q 53
Q 54
Q 55
Q 56
Q 57
ζ
ε57
ε58
ε62
ε64
ε63
ε54
ε55
ε56
ε50
ε51
ε52
ε53
Q 65ε65
Q 58
ε61
P1
P2
P3
Source : Author’s own compilation
In a first step propositions are formulated. Cooper and Schindler (2008:64) state that
propositions are statements about concepts that may be true or false. In chapter five
propositions are formulated for empirical testing and hence become hypotheses.
The following propositions derive from the measurement model:
P1: Market driving can be measured by market-sensing activities.
P2: Market driving can be measured by activities related to influencing customer
preferences.
P3: Market driving can be measured by alliance-formation activities.
- 171 -
4.4.1.2 Measurement model for corporate entrepreneurial management
Corporate entrepreneurial management is considered to be a second-order formed
abstract object consisting of three dimensions: risk-taking, management support and
organisational structure.
The dimensions consist of first-order eliciting attributes which are represented in
questions 1-10. Causality flows from the dimension, for example from risk-taking to
its items.
An index for corporate entrepreneurial management was developed which consisted
of the sum of item scores.
Risk-taking was measured by the two items (questions 9-10) developed by Miller and
Friesen (1982) which have been used in numerous previous studies (Kreiser et al.,
2002; Miles & Arnold, 1991; Morris & Sexton, 1996; Smart & Conant, 1994).
Management support consisted of four items (questions 5-8), which were derived
from Hornsby et al. (2002).
Organisational structure consisted of four items (questions 1-4), adapted from
Hornsby et al. (2002) and Khandwalla’s study (1977).
The following figure summarises the measurement model for corporate
entrepreneurial management.
- 172 -
FIGURE 4.3: Measurement model for corporate entrepreneurial management
RISK
MGT
STRU
CE
Q 9
Q 10
Q 5
Q 6
Q 7
Q 8
Q 1
Q 2
Q 3
Q 4CE Corporate entrepreneurial managementRISK Risk-takingMGT Management supportSTRU Organisational structure
ε Measurement error (reflective)ζ Disturbance term (formative)
ζ
ε6
ε7
ε8
ε1
ε3
ε2
ε9
ε10
ε5
ε4
P4
P5
P6
Source : Author’s own compilation
The propositions for corporate entrepreneurial management are as follows:
P4: Corporate entrepreneurial management can be measured by risk-taking
activities.
P5: Corporate entrepreneurial management can be measured by management
support.
P6: Corporate entrepreneurial management can be measured by organisational
structure.
4.4.1.3 Measurement model for entrepreneurial capital
Entrepreneurial capital is considered to be a second-order eliciting attribute which is
reflected in the three dimensions of financial, social and human capital. These three
- 173 -
dimensions are first-order eliciting attributes which consist of three items each. A total
item score for entrepreneurial capital was generated.
Financial capital was measured by three items (questions 35-37). The items were
self-constructed; ideas were taken from Miller and Friesen (1982) and Khandwalla
(1977).
Human capital included three items (questions 41-43) which were self-constructed;
ideas were taken from Unger et al. (2011) and Rauch et al. (2005).
Social capital consisted of three items (questions 38-40) which were self-constructed;
ideas for the item development were taken from Baron and Markman (2000).
The following figure summarises the measurement model for corporate
entrepreneurial management.
FIGURE 4.4: Measurement model for entrepreneurial capital
FIN
SOC
CA
Q 35
Q 36
Q 38
Q 39
Q 40CA Entrepreneurial capitalFIN Financial capitalSOC Social capitalHUM Human capital
ε Measurement error
Q 37
HUMQ 41
Q 42
Q 43
ε35
ε36
ε37
ε38
ε39
ε40
ε41
ε42
ε1
ε3
ε2
ε43
P7
P8
P9
Source : Author’s own compilation
- 174 -
The propositions for entrepreneurial capital are as follows:
P7: Entrepreneurial capital is reflected in financial capital.
P8: Entrepreneurial capital is reflected in human capital.
P9: Entrepreneurial capital is reflected in social capital.
4.4.1.4 Measurement model for strategic orientation
Strategic orientation is considered to be a second-order formed attribute which is
caused by four dimensions information generation, information dissemination,
interfunctional coordination and innovation intensity. These four dimensions are first-
order eliciting attributes. An index for strategic orientation was developed which
consisted of the sum of item scores.
Information generation was measured by four items (questions 17-20) from the scale
developed by Jaworski and Kohli (1993).
Information dissemination consisted of four items (questions 21-24) which were
adapted from Jaworski and Kohli (1993).
Interfunctional coordination consisted of four items (questions 25-28). The items were
derived from the scale developed by Narver and Slater (1990).
Innovation intensity consisted of three items (questions 29-31) which were adapted
from Miller and Friesen (1982).
- 175 -
FIGURE 4.5: Measurement model for strategic orientation
GEN
DIS
COO
SO
Q 21
Q 22
Q 23
Q 24
Q 25
Q 26
Q 27
Q 28
SO Strategic orientationGEN Information generationDIS Information disseminationCOO Interfunctional coordinationINN Innovation intensity
ε Measurement error (reflective)ζ Disturbance term (formative)
INN
Q 29
Q 30
Q 31
Q 17
Q 18
Q 19
Q 20
ζ
ε20
ε21
ε22
ε23
ε25
ε24
ε17
ε18
ε19
ε26
ε27
ε28
ε29
ε30
ε31
P10
P11
P12
P13
Source: Author’s own compilation
The propositions for strategic orientation are as follows:
P10: Strategic orientation can be measured by information generation.
P11: Strategic orientation can be measured by information dissemination.
P12: Strategic orientation can be measured by interfunctional coordination.
P13: Strategic orientation can be measured by innovation intensity.
4.4.1.5 Measurement model for entrepreneurial behaviour
Entrepreneurial behaviour is considered to be a second-order formed attribute which
is caused by the two dimensions of proactiveness and responsiveness to information.
These two dimensions are first-order eliciting attributes.
- 176 -
An index for entrepreneurial behaviour was developed which consisted of the sum of
item scores.
Proactiveness was measured by three items (questions 11-13); items were adapted
from Lumpkin and Dess (2001).
The three items used to measure responsiveness to information (questions 14-16)
were adapted from Jaworski and Kohli (1993) and Kohli et al. (1993).
FIGURE 4.6: Measurement model for entrepreneurial behaviour
PRO
RESP
BE
Q 11
Q 12
Q 14
Q 15
Q 16
BE Entrepreneurial behaviourPRO ProactivenessRESP Responsiveness to information
ε Measurement error (reflective)ζ Disturbance term (formative
Q 13
ζε11
ε12
ε13
ε14
ε15
ε16
P14
P15
Source : Author’s own compilation
The propositions for entrepreneurial behaviour are as follows:
P14: Entrepreneurial behaviour can be measured by proactiveness.
P15: Entrepreneurial behaviour can be measured by responsiveness to information.
- 177 -
4.4.1.6 Measurement model for firm performance and relative competitive
strength
Firm performance and relative competitive strength are considered to be first-order
eliciting attributes.
A summed score for each concept, firm performance and relative competitive
strength, was developed, which consisted of the sum of its item scores.
Firm performance was measured with three self-constructed items (questions 1A-1C)
that captured the perception of respondents on firm performance.
Relative competitive strength was measured with five items (questions 66-70), which
were adapted from Burke (1984).
FIGURE 4.7: Measurement model of firm performance and relative competitive
strength
Q 1A
Q 1B
PERF Firm performanceCOMP Relative competitive strength
ε Measurement error (reflective)
Q 1C
Q 66
Q 67
Q 68
Q 69
Q 70
ε1A
ε1B
ε1C
ε66
ε67
ε68
ε69
ε70
PERF
COMP
Source : Author’s own compilation
- 178 -
Firm performance and relative competitive strength are outcomes parameters of this
study and were hence tested as a part of the structural model in this study.
4.4.2 Structural models
The following section outlines three different hypothesised structural models. The first
model investigates a direct effects relationship between the exogenous latent
variables and market-driving ability. The second model determines whether the
management level as a moderating variable influences the paths between the
exogenous latent variables and market driving. In a third model the industry focus of
the organisation is considered as a moderating variable.
4.4.2.1 Direct effects model (model 1)
It is suggested that the structural model of market driving is a recursive model with
linear relationships and uncorrelated disturbance terms.
The exogenous latent variables in the model are corporate entrepreneurial
management (CE), entrepreneurial capital (CA), strategic orientation (SO) and
entrepreneurial behaviour (BE). The endogenous latent variables are market-driving
ability (MD-ability), firm performance (PERF) and relative competitive strength
(COMP).
The first structural model (Figure 4.8) presents direct relationships between the latent
variables. The double-headed curved arrows between the exogenous latent variables
represent an unanalysed relationship. It is acknowledged that the latent variables
could be correlated; however, this study is not going to explore if and why they are
correlated (Jaccard & Jacoby, 2010:143).
- 179 -
FIGURE 4.8: Direct effects model (model 1)
M-Dability
PERF
COMP
+
+
P20
P21
SO4 dimensions
2 dimensions
CE3 dimensions
BE
CA3 dimensions
3 dimensionsP16
P17
P18
P19
+
+
+
+
Source : Author’s own compilation
The propositions that derive from the path diagram are stated as follows:
P16: Corporate entrepreneurial management positively influences market-driving
ability.
P17: Entrepreneurial capital positively influences market-driving ability.
P18: Strategic orientation positively influences market-driving ability.
P19: Entrepreneurial behaviour positively influences market-driving ability.
P20: Market-driving ability positively influences firm performance.
P21: Market-driving ability positively influences relative competitive strength.
- 180 -
4.4.2.2 Moderating effects model: Management level (model 2)
In the second and third structural models, moderating effects are considered. A
moderator is a qualitative or quantitative variable that affects the direction and/or
strength of the relationship between an independent and a dependent latent variable
(Baron & Kenny, 1986:1174; Helm, Eggert & Garnefeld, 2010:524; Henseler &
Fassott, 2010:713).
The structural model considers the influence of management level (top, middle,
junior) on the relationship between the exogenous latent variables and market-driving
ability.
As previous research (Hornsby et al., 2002:260,269) found, upper middle managers
perceive key firm internal factors in a different way from lower middle managers.
Middle managers’ perception of these factors influenced their participation in
entrepreneurial endeavours. Therefore it would be interesting to see if a moderating
relationship could be found in this study.
For the purpose of this study it was investigated whether the path between the
exogenous latent constructs and market driving was influenced by the management
level. The specific question that was asked was: “Will the relationship between
corporate entrepreneurial management, entrepreneurial capital, strategic orientation,
entrepreneurial behaviour and market-driving ability be influenced by the various
management levels?”
The direction of the moderating effect was not hypothesised.
The following propositions derive from the second model:
P22: The path between corporate entrepreneurial management and market-driving
ability will differ between various levels of management.
P22a: The path between corporate entrepreneurial management and
market-driving ability will differ between top management (level 1) and
middle management (level 2).
- 181 -
P22b: The path between corporate entrepreneurial management and
market-driving ability will differ between middle management (level 2)
and junior management (level 3).
P22c: The path between corporate entrepreneurial management and
market-driving ability will differ between top management (level 1) and
junior management (level 3).
P23: The path between entrepreneurial capital and market-driving ability will differ
between various levels of management.
P23a: The path between entrepreneurial capital and market-driving ability will
differ between top management (level 1) and middle management
(level 2).
P23b: The path between entrepreneurial capital and market-driving ability will
differ between middle management (level 2) and junior management
(level 3).
P23c: The path between entrepreneurial capital and market-driving ability will
differ between top management (level 1) and junior management
(level 3).
P24: The path between strategic orientation and market-driving ability will differ
between various levels of management.
P24a: The path between strategic orientation and market-driving ability will
differ between top management (level 1) and middle management
(level 2).
P24b: The path between strategic orientation and market-driving ability will
differ between middle management (level 2) and junior management
(level 3).
- 182 -
P24c: The path between strategic orientation and market-driving ability will
differ between top management (level1) and junior management
(level 3).
P25: The path between entrepreneurial behaviour and market-driving ability will
differ for various management levels.
P25a: The path between entrepreneurial behaviour and market-driving ability
will differ between top management (level 1) and middle management
(level 2).
P25b: The path between entrepreneurial behaviour and market-driving ability
will differ between middle management (level 2) and junior
management (level 3).
P25c: The path between entrepreneurial behaviour and market-driving ability
will differ between top management (level 1) and junior management
(level 3).
- 183 -
FIGURE 4.9: Moderating effects model: Management level (model 2)
M-Dability
PERF
COMP
SO4 dimensions
2 dimensions
CE3 dimensions
BE
CA3 dimensions
3 dimensions
Management level
P22
P23
P24
P25
Source : Author’s own compilation
4.4.2.3 Moderating effects model: Industry focus (model 3)
The third structural model considers the moderating influence of industry focus of
organisations (Pharmaceutical manufacturers, medical device manufacturers,
pharmaceutical distributors/wholesalers and medical schemes) on the relationship
between the latent exogenous variables and market-driving ability. The industry focus
of organisations is associated with certain challenges in the respective market.
Previous research (Khandwalla, 1976/77; Zahra & Covin, 1995) found that events in
the external environment such as turbulence and dynamism in the market, restrictive
access to the market and technological sophistication influenced firms’ activities.
Zahra & Covin (1995:52) formed an environmental hostility index which took account
of profit margins, growth rates and the number of bankruptcies in order to determine
the relationship between corporate entrepreneurship and performance. These
- 184 -
quantitative data are not available for the different business sectors in the South
African healthcare industry included in this study. Therefore the influence of industry
focus on the paths was determined on a qualitative basis, considering the information
that could be obtained by secondary research for the individual business sectors.
The pharmaceutical industry, consisting of manufacturers of originals and generics, is
growing at a steady pace with prosperous growth expectations. In 2009 the total
pharmaceutical market was estimated at 19.8 billion South African Rand (ZAR),
which was a 15% increase on that of 2008 (Aspen Pharma, 2009:3). In 2010 the
market was expected to reach 24.56 billion ZAR and by 2014, 35.97 billion ZAR
(ReportLinker, 2011).
South Africa’s leading pharmaceutical organisations are constantly looking for
strategic alliances within the country as well as exploring expansion opportunities into
neighbouring African countries (Aspen Pharma, 2009:3; ReportLinker, 2011). The
pharmaceutical market will be influenced by patent expirations which will further
strengthen the position of generic organisations (ReportLinker, 2011). Furthermore,
government’s efforts for a national health insurance system will put pressure on the
industry for lower cost and increased patient access to medication.
These expectations about market development lead to the hypothesis that
pharmaceutical organisations have a strong need to increase their market-driving
ability, as the industry is considered highly competitive, with various challenges that
impact on growth opportunities and market access.
The South African medical device industry is considered to be dynamic and highly
competitive, with national and international players. It is estimated that approximately
95% of devices are imported (Anon., 2010:3). Market growth in the medical device
industry is expected at 7.1% from 2010-2015 (Episcom Healthcare Intelligence,
2010). Currently the medical equipment market is not well regulated as there is no
comprehensive medical device regulation system (Medicaldevice-network.com,
2009). The positive growth outlook can be considered to enhance firm’s activity in
market driving. However, the uncertainty regarding the regulatory environment could
provide obstacles to market-driving endeavours.
- 185 -
Two of the leading distributors of pharmaceuticals are experiencing continued growth
of their business. Innovations regarding distribution systems that increase turnaround
times, provide process efficiency and add flexibility to the client, are being introduced.
However, a regulation by the government to cap logistical fees provides future
challenges to pharmaceutical distributors (UTI Pharma, 2011). Considering this
information, the distributors of pharmaceuticals will have a strong need to achieve
market driving for their organisations in order to tackle the challenges ahead and
ensure continuing success.
The medical schemes environment in South Africa is characterised by
consolidations, liquidations and mergers. The number of medical schemes is
constantly declining. From 2006 to 2009 the number of open medical schemes
decreased by almost 50%, from 218 to 110 schemes. Currently 33 open medical
schemes operate in South Africa. The number of principal members increased
steadily over the past years with an increase rate of 4.8% for the period 2006-2007
and 2.9% for the period 2008-2009. However, more and more elderly people join
medical schemes, with an increase of 0.4% and 0.6% for the periods 2006-2007 and
2008-2009 respectively. Furthermore, spending on hospitals, medical specialists and
medicines accounts for the majority of total spending, which increased by 13.7% from
2006 to 2007 and by 18% from 2008 to 2009. (Council of Medical Schemes,
2009:123-125; 2010:157-165).
Based on this information it can be assumed that medical schemes need to increase
their activities towards market driving in order to stay competitive and tackle the
challenges of rising healthcare costs and an ageing member portfolio. However, it
can also be assumed that, based on the rising cost and the aspirations of the
government to introduce a national healthcare system, medical schemes will resign
and will not engage in market-driving activities.
It is hypothesised that the industry focus will impact on the paths between the
dimensions and market driving. Considering the opportunities and challenges of each
industry sector, the direction of the effect cannot be specified.
- 186 -
P26: The path between corporate entrepreneurial management and market-driving
ability will differ for various industries.
P26a: The path between corporate entrepreneurial management and
market-driving ability will differ between pharmaceutical manufacturers
and medical device manufacturers.
P26b: The path between corporate entrepreneurial management and
market-driving ability will differ between medical device manufacturers
and pharmaceutical distributors/wholesalers.
P26c: The path between corporate entrepreneurial management and
market-driving ability will differ between pharmaceutical manufacturers
and pharmaceutical distributors/wholesalers.
P26d: The path between corporate entrepreneurial management and
market-driving ability will differ between pharmaceutical manufacturers
and medical schemes.
P26e: The path between corporate entrepreneurial management and
market-driving ability will differ between medical device manufacturers
and medical schemes.
P26f: The path between corporate entrepreneurial management and
market-driving ability will differ between pharmaceutical
distributors/wholesalers and medical schemes.
P27: The path between entrepreneurial capital and market-driving ability will differ
for various industries.
P27a: The path between entrepreneurial capital and market-driving ability will
differ between pharmaceutical manufacturers and medical device
manufacturers.
- 187 -
P27b: The path between entrepreneurial capital and market-driving ability will
differ between medical device manufacturers and pharmaceutical
distributors/wholesalers.
P27c: The path between entrepreneurial capital and market-driving ability will
differ between pharmaceutical manufacturers and pharmaceutical
distributors/wholesalers.
P27d: The path between entrepreneurial capital and market-driving ability will
differ between pharmaceutical manufacturers and medical schemes.
P27e: The path between entrepreneurial capital and market-driving ability will
differ between medical device manufacturers and medical schemes.
P27f: The path between entrepreneurial capital and market-driving ability will
differ between pharmaceutical distributors/wholesalers and medical
schemes.
P28: The path between strategic orientation and market-driving ability will differ for
various industries.
P28a: The path between strategic orientation and market-driving ability will
differ between pharmaceutical manufacturers and medical device
manufacturers.
P28b: The path between strategic orientation and market-driving ability will
differ between medical device manufacturers and pharmaceutical
distributors/wholesalers.
P28c: The path between strategic orientation and market-driving ability will
differ between pharmaceutical manufacturers and pharmaceutical
distributors/wholesalers.
- 188 -
P28d: The path between strategic orientation and market-driving ability will
differ between pharmaceutical manufacturers and medical schemes.
P28e: The path between strategic orientation and market-driving ability will
differ between medical device manufacturers and medical schemes.
P28f: The path between strategic orientation and market-driving ability will
differ between pharmaceutical distributors/wholesalers and medical
schemes.
P29: The path between entrepreneurial behaviour and market-driving ability will
differ for various industries.
P29a: The path between entrepreneurial behaviour and market-driving ability
will differ between pharmaceutical manufacturers and medical device
manufacturers.
P29b: The path between entrepreneurial behaviour and market-driving ability
will differ between medical device manufacturers and pharmaceutical
distributors/wholesalers.
P29c: The path between entrepreneurial behaviour and market-driving ability
will differ between pharmaceutical manufacturers and pharmaceutical
distributors/wholesalers.
P29d: The path between entrepreneurial behaviour and market-driving ability
will differ between pharmaceutical manufacturers and medical
schemes.
P29e: The path between entrepreneurial behaviour and market-driving ability
will differ between medical device manufacturers and medical
schemes.
- 189 -
P29f: The path between entrepreneurial behaviour and market-driving ability
will differ between pharmaceutical distributors/wholesalers and
medical schemes.
FIGURE 4.10: Moderating effects model: Industry focus (model 3)
M-Dability
PERF
COMP
SO4 dimensions
2 dimensions
CE3 dimensions
BE
CA3 dimensions
3 dimensions
Industry focus
P26
P27
P28
P29
Source: Author’s own compilation
4.5. CONCLUSION
The chapter provided a literature overview relating to measuring instruments that
have been used in past research and are relevant for this study. The instruments’
dimensions and their reported reliability and validity measures were presented.
Relevant literature on statistical modelling was also presented. It was explained that
the two parts of the model, the measurement model and the structural model, need to
be specified in order to achieve complete model specification. The measurement
model, which is certainly the more difficult part to specify, can consist of reflective or
- 190 -
formative relationships between a construct and its dimensions and the dimensions
with their indicators. The impact of model misspecification was also discussed.
The third part of this chapter addressed the development of a model for
market-driving ability in corporate entrepreneurship. In a first step the measurement
model for each construct was developed. Measurement model specification followed
the guidelines outlined by Rossiter (2002). The constructs of market-driving,
corporate entrepreneurial management, strategic orientation, and entrepreneurial
behaviour were defined with reflective first-order indicators and formative second-
order dimensions. Entrepreneurial capital was defined with reflective first-order
indicators and reflective second-order dimensions. Performance and relative
competitive advantage were considered to consist of first-order reflective indicators.
In a second step the structural models for this study were specified. The first model
represents a direct effects model and considers that the exogenous constructs:
corporate entrepreneurial management, entrepreneurial capital, strategic orientation
and entrepreneurial behaviour, directly and positively influence market-driving ability.
It was also specified that market-driving ability positively influences relative
competitive strength and firm performance.
The second model considers the moderating effect of the management level on the
structural relationships. The third model considers the moderating effect of industry
focus on the relationships between the exogenous latent constructs and
market-driving ability.
The chapter concluded with the explicit specification of propositions that derive from
the specified path diagrams.
Top Related